CVE-2026-91776: 
Java vulnerability analysis and mitigation

Overview

CVE-2026-91776 is an Uncontrolled Resource Consumption (CWE-400) vulnerability in FasterXML jackson-databind, titled "jackson-databind retains every unknown raw type ID." When name-based polymorphic deserialization is configured with a fallback (e.g., @JsonTypeInfo(use = Id.NAME, defaultImpl = ...)), the TypeDeserializerBase._findDeserializer() method caches each distinct attacker-supplied unknown type ID as a separate key in an unbounded _deserializers map, enabling monotonic heap growth and eventual denial of service. Affected versions include com.fasterxml.jackson.core:jackson-databind 2.0.0–2.18.10, 2.19.0–2.21.6, and 2.22.0–2.22.2, as well as tools.jackson.core:jackson-databind 3.0.0–3.1.6 and 3.2.0–3.2.2. The vulnerability was reported by Daniel Birtwhistle, published on September 22–23, 2026, and carries a CVSS v3.1 base score of 7.5 (High) (GitHub Advisory, Jackson PR #6203).

Technical details

The root cause is a logic flaw (CWE-400) in TypeDeserializerBase._findDeserializer(): after an unknown name-based type ID resolves to the configured defaultImpl fallback deserializer, jackson-databind caches the result under the raw, attacker-supplied type ID string as a distinct key in the _deserializers map. Because all unknown IDs resolve to the same fallback, the cached entries provide no lookup benefit, yet each unique string permanently occupies a map slot. The map has no configurable size bound and lives for the lifetime of the TypeDeserializer (typically the application lifetime when a shared ObjectMapper is used). Exploitation requires three preconditions: (1) name-based polymorphism configured with defaultImpl or equivalent fallback, (2) attacker-controlled type ID strings in JSON input, and (3) a long-lived, shared ObjectMapper instance reused across requests. A PoC demonstrating 10,000 distinct unknown type IDs producing 10,000 retained cache entries (versus one entry when the same unknown ID is repeated 10,000 times) is described in the official advisory (GitHub Advisory, Jackson PR #6203).

Impact

Successful exploitation causes monotonic JVM heap growth proportional to the number of distinct unknown type IDs submitted by an attacker, ultimately leading to OutOfMemoryError, application crash, or severe performance degradation — a denial of service. There is no confidentiality or integrity impact; code execution is not possible through this vulnerability. Downstream products incorporating jackson-databind, including IBM DevOps Build (Rational Build Forge) and IBM InfoSphere Data Architect 9.2.2, are also affected (GitHub Advisory, IBM Advisory).

Exploitability

A proof-of-concept is publicly available in the official GitHub Security Advisory, which provides concrete reproduction steps (GitHub Advisory). The vulnerability is automatable (no authentication or user interaction required) and exploitable over the network with low attack complexity. The EPSS score is approximately 0.45–0.49%, indicating a relatively low but non-negligible probability of exploitation in the wild within 30 days. There is no evidence of active in-the-wild exploitation at this time, and the CVE is not listed in the CISA KEV catalog (GitHub Advisory, Feedly).

Exploitation steps

  1. Reconnaissance: Identify applications that use a vulnerable version of jackson-databind (2.0.0–2.18.10, 2.19.0–2.21.6, 2.22.0–2.22.2, 3.0.0–3.1.6, or 3.2.0–3.2.2) and expose a JSON deserialization endpoint where polymorphic type information is accepted from user input.
  2. Confirm vulnerable configuration: Verify that the target application uses @JsonTypeInfo(use = JsonTypeInfo.Id.NAME, defaultImpl = SomeFallback.class) or an equivalent fallback configuration, and that the same ObjectMapper instance is reused across requests.
  3. Craft JSON payload with a unique unknown type ID: Construct a JSON object containing a type field with an arbitrary string that does not match any registered polymorphic subtype name, e.g., {"@type": "unknown_type_1", ...}.
  4. Send payload to the vulnerable endpoint: POST the crafted JSON to the application's deserialization endpoint via HTTP.
  5. Repeat with distinct unknown type IDs: Automate sending thousands of requests, each with a unique unknown type ID string (e.g., unknown_type_1 through unknown_type_10000), maximizing cache key cardinality in the _deserializers map.
  6. Trigger unbounded memory growth: Each distinct type ID is cached as a separate key in TypeDeserializerBase._deserializers, even though all resolve to the same fallback deserializer. The map grows without bound across the application's lifetime.
  7. Achieve denial of service: Sufficient accumulation of cache entries exhausts JVM heap memory, causing OutOfMemoryError, application crash, or severe performance degradation (GitHub Advisory, Jackson PR #6203).

Indicators of compromise

  • Network: High volume of HTTP POST requests to JSON deserialization endpoints, each with a distinct, unrecognized value in the polymorphic type field (e.g., @type, type, or a custom type property); requests may appear structurally similar but differ only in the type ID string value.
  • Application Logs: Repeated deserialization of unknown type IDs that fall back to a default implementation, potentially logged as warnings or debug messages by jackson-databind or the application framework; absence of errors despite many unknown type names may indicate silent fallback caching.
  • JVM / Process Metrics: Monotonically increasing JVM heap usage over time without corresponding increase in legitimate application load; frequent or prolonged garbage collection cycles that fail to reclaim memory; eventual java.lang.OutOfMemoryError in application logs or JVM crash dumps.
  • Heap Dumps: Presence of a large TypeDeserializerBase._deserializers map (e.g., thousands of entries) in a JVM heap dump, with keys consisting of arbitrary or random-looking strings and all values pointing to the same fallback deserializer class (GitHub Advisory).

Mitigation and workarounds

Upgrade to the following patched versions of jackson-databind: 2.18.11, 2.21.7, 2.22.3, 3.1.7, or 3.2.3. The fix stops caching fallback resolutions for unrecognized type IDs, caps the _deserializers map at 1,000 entries (clearing on overflow), and limits cacheable type ID length to 256 characters. As interim mitigations, avoid exposing polymorphic deserialization endpoints with defaultImpl fallbacks to untrusted input, restrict the scope of shared ObjectMapper instances, and implement request-level input validation to reject excessively long or high-cardinality type ID fields. Monitor JVM heap usage for anomalous growth in applications using name-based polymorphism (GitHub Advisory, Jackson PR #6203, IBM Advisory).

Community reactions

The fix was developed by contributor pjfanning and merged by cowtowncoder (FasterXML maintainer) on September 14, 2026, with patches released across five version branches (Jackson PR #6203). The Apache Spark Kubernetes Operator project promptly upgraded to Jackson 2.22.3 to address this and related CVEs, and the Quarkus project released 3.33.4 incorporating the fix. The parallel-consumer project temporarily excluded the vulnerable jackson-databind version with a documented expiry, reflecting broad community awareness and active dependency management in response to the advisory. SUSE issued a security update (SUSE-SU-2026:4394-1) and openSUSE published a corresponding security announcement, indicating distribution-level uptake of the fix (SUSE Advisory).

Additional resources

Linux Distribution fix status

Fix availability across major Linux distributions and their releases.

Debian

Fixed

bookworm

jackson-databind

Affected

sid

jackson-databind: 2.18.11-1

Fixed

trixie

jackson-databind

Affected

Ubuntu

Affected

bionic (esm-apps)

jackson-databind

Affected

devel

jackson-databind

Unknown

focal (esm-apps)

jackson-databind

Unknown

jammy

jackson-databind

Unknown

jammy (esm-apps)

jackson-databind

Unknown

noble

jackson-databind

Affected

noble (esm-apps)

jackson-databind

Affected

resolute

jackson-databind

Unknown

RHEL / CentOS

Affected

RHEL 8

Not Affected

RHEL 9

Not Affected

RHEL 10

Not Affected

Source: This report was generated using AI

Related Java vulnerabilities:

CVE ID

Severity

Score

Technologies

Component name

CISA KEV exploit

Has fix

Published date

CVE-2026-106554HIGH7.5
  • Java logoJava
  • pki-deps:10.6::jackson-jaxrs-json-provider
NoYesNov 04, 2026
CVE-2026-46487HIGH7.5
  • Java logoJava
  • org.geonetwork-opensource:geonetwork
NoYesNov 03, 2026
CVE-2026-75598MEDIUM6.3
  • Java logoJava
  • com.gradleup.shadow:com.gradleup.shadow.gradle.plugin
NoYesNov 06, 2026
CVE-2026-59171MEDIUM5.9
  • Java logoJava
  • firefox
NoYesNov 06, 2026
CVE-2026-54442MEDIUM4.8
  • Java logoJava
  • gov.nsa.emissary:emissary
NoYesNov 05, 2026

Free Vulnerability Assessment

Benchmark your Cloud Security Posture

Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.

Request assessment

Additional Wiz resources

Get a personalized demo

Ready to see Wiz in action?

"Best User Experience I have ever seen, provides full visibility to cloud workloads."
David EstlickCISO
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
Adam FletcherChief Security Officer
"We know that if Wiz identifies something as critical, it actually is."
Greg PoniatowskiHead of Threat and Vulnerability Management